Hybrid Group Anomaly Detection for Sequence Data: Application to Trajectory Data Analytics

Many research areas depend on group anomaly detection. The use of detection can maintain and provide security privacy to the data involved. This attempts solve deficiency existing literature in outlier thus a novel hybrid framework identify from sequence is proposed this paper. It proposes two approaches for efficiently solving problem: i) Hybrid Data Mining-based algorithm , consists three main phases: first, clustering algorithm applied derive micro-clusters. Second, $kNN$ each micro-cluster calculate candidates group’s outliers. Third, pattern mining gets outliers as pruning strategy, generate groups outliers, ii) xmlns:xlink="http://www.w3.org/1999/xlink">GPU-based approach presented, which benefits massively GPU computing boost runtime mining-based algorithm. Extensive experiments were conducted show advantages different databases our model. Results clearly efficiency direction when directly compared sequential by reaching speedup 451 . In addition, both outperform baseline methods